Fwd: Re: film at 11: kernel update breaks udev.
Bill Nottingham
notting at redhat.com
Mon Jul 23 14:16:27 UTC 2007
Harald Hoyer (harald at redhat.com) said:
> What makes you think so?
>
> Is the udevd binary slower?
>
> Is selinux making it slower?
>
> Are the default rules in
> /etc/udev/rules.d/05-udev-early.rules
> /etc/udev/rules.d/50-udev.rules
> /etc/udev/rules.d/95-pam-console.rules
> slow?
>
> What is upstream? udev-113/etc/udev/redhat or udev-113/etc/udev/suse ????
I've never noticed it being 100x slower. However, there are a couple of
cases for improvement:
- make_extra_nodes in start_udev is relatively slow. The more we can avoid
using this, the better
- 60-libsane.rules takes measureable time in udevd just to *parse*, even though
they're skipped on bootup. As part of deprecating/removing pam_console
(http://fedoraproject.org/wiki/Releases/FeatureRemovePAMConsole), these
can hopefully go away
- the bluetooth rules could use some skip/goto
- is check_cdrom.sh really needed?
But still, this only shaves startup by about a second or so.
Bill
More information about the devel
mailing list